[PATCH 04 of 10 py3] scmutil: convert exception to bytes in a Py3-friendly manner

[Augie Fackler]

> On May 30, 2017, at 08:49, Yuya Nishihara <yuya at tcha.org> wrote:
> 
> On Mon, 29 May 2017 10:32:26 -0400, Augie Fackler wrote:
>> # HG changeset patch
>> # User Augie Fackler <raf at durin42.com>
>> # Date 1496000686 14400
>> #      Sun May 28 15:44:46 2017 -0400
>> # Node ID 6e1ab7424617c7c7afca7faab559507c5e816a9c
>> # Parent  f1ae981927af4d2d75e2a0bc58c733c9a90c085b
>> scmutil: convert exception to bytes in a Py3-friendly manner
>> 
>> As far as I can tell, the Abort instance might contain localized data,
>> so we need to str() then convert to bytes by hand.
>> 
>> diff --git a/mercurial/scmutil.py b/mercurial/scmutil.py
>> --- a/mercurial/scmutil.py
>> +++ b/mercurial/scmutil.py
>> @@ -187,7 +187,10 @@ def callcatch(ui, func):
>>             ui.warn(_("(%s)\n") % inst.hint)
>>         return 1
>>     except error.Abort as inst:
>> -        ui.warn(_("abort: %s\n") % inst)
>> +        msg = str(inst)
>> +        if not isinstance(msg, bytes):
>> +            msg = encoding.unitolocal(msg)
>> +        ui.warn(_("abort: %s\n") % msg)
> 
> This works as follows on Python 3:
> 
> 1. str(inst) converts non-ascii bytes in b'\xXX' form
> 2. escaped "b'...'" unicode is converted back to bytes
> 
> Perhaps we'll need to add __bytes__() to our exception classes.

Indeed, I'd believe that. Should we proceed with this patch as-is and revisit the particular b'' repr ugly when we trip on it?

(I'm honestly not sure how we should proceed here.)